|zpravodaj ČSKI - květen 2007 [ pdf ]|
datum: 30.5.2007 v 14:00
název: Locating Landmarks Using Templates
přednášející: Jan Kalina (MFF UK)
místo konání: UTIA, 4.patro, místnost č.474
souhrn: Our work is devoted to automatic location of landmarks (mouth and eyes) in 2Dimages of faces using templates. The experience with existing software is unsatisfactory because of its high sensitivity to small rotations of the face. The weighted correlation coefficient as a similarity measure between the template and the image turns out to outperform the classical sample correlation. It is presented how to choose the weights to increase the discrimination of the parts of the face, which correspond to the template from those, which do not. Optimization without constraints tends to degenerate and to obtain a robust version we bound the influence of single pixels. In a similar way the template can be optimized to improve the discrimination further. The results are compared for different initial choices of weights and their robustness to different size or rotation of the face is examined. The method does not use any special properties of the mouth or eyes and can be classified as a robust nonparametric discrimination technique.